Request for correction: MACE manifest draft in OpenSciFlow

[kkkniengLiu]

Hi MACE maintainers,

I am maintaining an early open initiative called OpenSciFlow: https://github.com/OpenSciFlow

We are drafting small, reviewable plugin manifests for AI-for-science tools and models so local/HPC workflows can record inputs, model weights, commands, outputs, citations, limitations, and run metadata before execution.

We created a draft MACE manifest here:

• Manifest: https://github.com/OpenSciFlow/plugin-manifest/blob/main/examples/mace-interatomic-potential/opensciflow.yaml
• Review notes: https://github.com/OpenSciFlow/plugin-manifest/blob/main/docs/model-manifest-review-notes.md#mace
• Landscape entry: https://github.com/OpenSciFlow/awesome-ai4s-workflows/blob/main/categories/scientific-models.md

This is not a request for support, adoption, or integration work. I am opening this only to ask whether the draft is accurate enough and whether anything is misleading.

Specific corrections that would help:

1. Is mace_eval_configs --configs ... --model ... --output ... the right minimal inference/evaluation command to represent?
2. Should each MACE foundation model family have a separate manifest?
3. Which model metadata fields should be required before a model is marked ready: elements, training domain, level of theory, target system, release, license, checksum?
4. How should OpenSciFlow represent per-model license differences?
5. Are the limitations around applicability domain and treating energies/forces as model predictions strong enough?
6. Is there a better citation or docs link we should use?

If Discussions are not the right place for this, please feel free to close it. A short correction comment is enough; I can make the change on the OpenSciFlow side.